This paper is the intellectual property of the author(s). It was presented at CAUSE98, an EDUCAUSE conference, and is part of that conference's online proceedings. See http://www.educause.edu/copyright.html for additional copyright information.

IP Telephony: The End Of The World As We Know It?
Scott V. Street
Senior Consultant
Compass Consulting International, Inc.
15 Brook Street
Medfield, MA 02052

Abstract

Industry pundits are touting the imminent demise of the PBX. The phone system, as the story goes, is to be replaced by a specialized telecommunications server that resides on the campus backbone, utilizes standard data communications protocols, and sends calls out over the "free" Internet. For information, data and voice professionals and their clients, we are told, "it's the end of the world as we know it."

Or is it?

This presentation will examine what is being said by telecommunications experts and industry players and present an overview of the conflicting arguments, likely outcomes, timeframes, and planning strategies.

Introduction

"This acquisition [of IP telephony supplier Selsius by Cisco] signals the new world in PBX technology. Customers should seriously consider the network PBX technology before committing to purchase old world equipment."

Mario Mazzola, Senior VP, Cisco Systems;

Cisco Systems to Acquire Selsius Systems, Inc.;

Press Release; October, 1998

It�s the end of the world as we know it.

We are hearing more and more that traditional voice technology, including the standard PBX, is going the way of the telegraph key and the Pony Express: valuable and functional communications systems being replaced by new, bigger, better, cheaper technologies. Voice and data networks, once separate and distinct, are merging into a single network; the PBX is dead or dying, to be replaced by networked telephony servers; computers are doubling as telephones. Management of this new technology environment is easier and more efficient, and these efficiencies translate into savings.

Further evidence of the revolution abounds. Telecommunications departments are now reporting to the same boss as their networking counterparts: the Chief Information Officer. Telecommunications vendors are gearing their sales pitches in order to win the "IT-influenced decision". Some PBXs run on NT servers; some PBX nodes are connected over ATM backbones. And our old friend, the Internet protocol (IP), is being combined with telecommunications technology to form a dynamic new hybrid: IP Telephony, a/k/a Voice Over IP (VoIP).

And best of all, thanks to IP telephony, phone calls are free! Wow!

Has your boss heard about this?

Digital Telephony

For purposes of this paper, we will refer to the above technologies generically as digital telephony, meaning voice systems that are computer-based and ride on the data network. Traditional voice communication has long been digitally based, including phones, switching, signaling, etc.; what we are discussing here is the newest applications of computer and network-based technologies for voice communication.

So once again, IT professionals are confronted with a familiar problem: how to separate the reality from the hype, i.e., the next Navigator from the next Newton, the next ethernet from the next arcnet. Is the revolution for real or simply a figment of an over-active media hype machine? Is the PBX really dead?

More specifically, given that this new technology will indeed play a part in the next generation of telecommunications equipment and services, which components and systems should IT leaders buy now, which should be implemented later, and which will never pan out? What will be the impact of digital telephony on IT budgets? What will happen to services being offered to students, faculty and staff? How will operations management, IT staffing and the IT department organizational structure be affected?

And don�t forget the end users�

This paper will offer a perspective on these questions. It will also examine ways in which IT professionals can understand the digital telephony "revolution" and its potential impact on their campus and their careers, and it will suggest strategies for dealing with the both the technology and the hype in the near and long term.

What�s The Buzz? Convergence

"If you look at the last 10 years and what�s happened in the network market, the Internet is growing at 1,000% per year and voice is growing at about 10%. I think six years from now, the voice market will be 2% of the network and the Internet will be 98%.

Piyush Patel; Senior VP of Engineering; Cabletron Systems
Investor�s Business Daily; October, 1998

The trend towards digital telephony is typically referred to by the industry as convergence. Convergence is the melding of voice, data, video, and multi-media technologies, formerly disparate entities, into a single, unified, digitally-based technology. This new technology encompasses:

Traditionally, voice is transmitted through a circuit switched network, with the caller being connected to the called party over a dedicated, one-to-one connection (a circuit). These connections are fast, reliable, cheap and, except in remote areas, ubiquitous. On the down side, each call requires a dedicated, point-to-point connection on a dedicated network, and bandwidth is limited.

In the new, converged world, a phone call is first packetized, either by a computer that connects the caller to the outside world (via a telephony server, for example) or by the telephone itself (e.g., an IP telephone). The packetized call is sent out as electronic data. This data can ride over a variety of ATM, SONET, and frame relay networks, not to mention over the Internet. The receiving end of the call, as with the sending end, must have the equipment necessary to either receive the call as a packetized signal via an IP device or have a telephony server that can de-packetize the call so it can be receive by a traditional phone.

Sounds easy enough�

The obvious advantage of the convergence approach is that instead of many things to buy, to install, to manage, to learn to use, and to pay for � cabling, connections, devices, tools � there is just one of each of these. Fewer moving parts means (theoretically) smoother operations, happier customers, easier management and lower costs. And since the voice call is utilizing unused bandwidth on the data network, voice essentially "rides for free" on the existing network infrastructure.

Did we mention that long distance calls over the Internet are free?

The Death Of The PBX

"�some computer telephony industry insiders predict [it] is the beginning of the end for PBXs."

Pedro Pereira;
Goodbye PBX: Data Market Is Seen As Ripe For Telephony Solutions;

Computer Reseller News, 27-Apr-98

"The PBX is dead!"

attributed to Bill Gates
Comdex, 1996

The above quotations notwithstanding, few industry pundits are unequivocally predicting the death of the PBX (and those who are may be somewhat biased � see sources above). There are many reasons for this unwillingness to jump completely on the "PBX is dead" bandwagon: for one thing, there is currently a very large, relatively satisfied PBX customer base holding substantial investments in equipment, infrastructure, and management. Managers who are happy with a sound, reliable technology investment are unlikely to jump at the chance to adopt the latest and greatest.

From a technical point of view, PBX technology is not so much being replaced as it is evolving into a leaner, more flexible, open system that can live on the data network. It is a transformation much like the one undergone by the mainframe computer of the 80s.

 

Remember The Main[Frame]?

Prior to the advent of the PC, mainframes were the way that computing was done on campus (and everywhere else). Mainframe computers were BIG and centralized, utilizing proprietary operating systems and cable plants; they were mission-specific, hardwired and inflexible; they were stand-alone systems accessed only via dumb terminals; they were very complex, monolithic, and expensive.

If you are familiar with the architecture and role of the basic PBX, the above characteristics should sound familiar. Like its "big iron" cousin, the traditional telephone system is proprietary, centralized, living on its own network, complex, monolithic and expensive to own and operate. It is also single-purpose, un-integrated with other systems, and inflexible.

PCs arrived big time in the mid-80s and have changed the computing paradigm from mainframe/dumb terminal to networked client/server. Yet rather than signaling the death of the mainframe, these large mainframe systems have evolved into good corporate citizens: they now live on the data network; they share data electronically with other systems; they are based on open, standards-based principles; and instead of a monolithic architecture, they are now scalable and designed so that smaller, distributed boxes can be coupled together.

Like the mainframe, the PBX is moving into the networked, open systems world. The PBX can now share the same cable infrastructure, databases, operating system, and client/server architecture, and even connections to the off-campus world that are utilized by the data network. What was once complex to manage is now done with a mouse click; what was centralized is increasingly being distributed across campus; what was unwieldy is now flexible. In short, the PBX, like the mainframe, is not sunk.

Not death by revolution. Adaptation by evolution.

Time To Make The Move To Digital Telephony?

The question remains when (and whether) to implement digital telephony in some form or other. It is all well and good to argue that standard voice technology is and will remain viable for the foreseeable future. Your customers (and your boss!) are seeing articles and ads touting free internet long distance calls, and they may soon demand to know what you are going to do about it, if they haven�t already done so. We like to refer to this phenomenon as "the peril of the in-flight magazine."

One wag calls it "reading edge technology."

IT decision-makers are often faced with this kind of question, and the criteria used to decide on implementing digital telephony are the same as those used for any technology decision. In our opinion, there are four major factors that drive the move to a new technology. These factors are:

In order to make a successful and, equally important, a justifiable technology change, you need economics and at least one other factor in support of the change.

And hype doesn�t count.

Our equation for a solid decision for technology is thus:

Economics + (Requirements and/or Manageability) � Hype =
Successful Technology Decision

Here is a look at the pros and cons for digital telephony for each of these factors.

Change Factors: Economics

A move to a new technology like IP telephony must be financially advantageous to the institution and/or your customers. A new system that is full of whiz-bang technology and makes your job easier typically will not convince the powers-that-be that they should buy it unless there is also some financial benefit for the institution (or is at least a break-even proposition). There are several economic factors to consider: saving on long distance calling (free calls!); utilizing one network instead of two; and the technology life cycle.

Free Calls

First, about those free calls. Yes, sending calls over the Internet avoids the cost of long distance access fees, accounting rates and settlement rates. Right now, the cost of a long distance telephone call is driven not so much by the actual cost to carry the call, but by charges levied by and on long distance providers and by what the market will bear. Internet-carried calls do not have to pay these charges � yet. However, according to Brett Azuma of Dataquest, "By 2002, access charges on traditional and Internet calls will even out at 2 cents per minute. These additional charges will remove Internet telephony�s most important source of cost and price advantage versus traditional services." (Lenore V. Tracey; "Voice over IP: Analyzing the Analysts;" Telecommunications Online; September 1998)

In other words, the arbitrage advantages of IP versus traditional calls is going away. When you couple this prediction with the ever-falling long distance rates, this conclusion is not surprising. Five cents per minute is becoming the standard; Sprint even offers calls for free! You can call Great Britain for 12 cents per minute on some plans. So while IP calls are cheap, their cost is likely to rise while the cost of traditional calls is falling. No long term economic advantage here.

One Network

Now take a look at the savings factor resulting from the use of one network instead of two. Lucent Technologies, in a recent press release (October 21, 1998), put it this way: "IP Exchange Systems are the next step in the evolution of communications. By making it easier for businesses to manage sophisticated voice and data communications over one network, the IP Exchange Systems deliver on Lucents promise to help customers reduce the cost and complexity of their networks while giving them the flexibility and power of an open systems platform." [italics added] It is interesting to note that while all of the big telecom players are making similar claims, most are shying away from touting the arbitrage cost advantages discussed above as a reason to change.

There is no question that data networks frequently have considerable unused capacity: According to an article in USA Today (October 8, 1998), it took 6 fibers to broadcast a single football game in 1985; today, one fiber can transmit the equivalent of 700 football games simultaneously, a 4,200-fold increase. Voice calls require very little bandwidth, especially at 10:1 compression ratios (projected to reach 30:1). If these calls are sent out over the unused bandwidth, then "voice rides for free" (more or less). Coupled with the notion that managing one network is easier and more efficient than managing two (and thus cheaper), you have the crux of the value proposition that Lucent and others are setting forth.

This is a strong argument, but it is perhaps not as strong as it might appear. The fact is that institutions typically already own a voice infrastructure, so moving off of it and onto the data side doesnt really save anything. In addition, implementing IP telephony requires a significant investment in new equipment (telephony servers, end user devices/software, sometimes cabling, switches, routers, etc.), not to mention the training, support, documentation and shake-down costs associated with any new technology. There are also the less obvious costs associated with staff and department reorganization, frustrated new users, and the general upheaval of the institutions culture. Nothings for nothing.

Technology Life Cycle: Bang for the Buck

The growth and development of any new technology follows a predictable life cycle, and understanding how far in the cycle a given technology like digital telephony has come is the best way to maximize technology investments (and minimize mistakes and headaches).

Clearly, the best time to gain the most bang for the buck is at the end of the specialty phase and into the commodity phase.

High value, low cost is good.

IP telephony and related technologies are relatively new on the scene and are still being developed. We see them in the early specialty system phase, moving out of the R&D phase; therefore it is too early to buy from a value standpoint, although not by much. It probably goes without saying that it is generally not prudent to be an early adopter of anything that is replacing something as familiar and reliable as the telephone.

Economics: The Bottom Line

In its coverage of the PBX 2000 conference (1998), Business Communications Review stated: "The economics of the new alternatives has yet to be proved." Further, Dr. Lee McKnight of MIT stated " in no uncertain terms that the business case for this hot technology is not ready for prime time." (Fred Knight, "Mood and Attitude Changes at PBX 2000; BCR; February 1998)

We concur. The economic benefits may be realized over the next five years, but not yet.

Change Factors: Functional/Technical Requirements

The need and/or desire for new and improved functionality is another factor that drives the acquisition of any new technology. Several technical advantages of digital telephony are frequently cited by its proponents, including:

And now, the rest of the story

While there are some legitimate functional and technical advantages provided by digital telephony, there remain some serious technical problems as well. Latency and reliability are the most often cited problems, and other major hurdles also must be overcome.

Latency

The latency problem will sound familiar to data network managers. Briefly stated, it takes time to digitize a telephone signal, compress it, "packetize" it for data network transmission, transmit it through any number of on- and off-campus switches and routers, and buffer it at the far end to make sure all the packets arrive before transmitting it to the end user. At the far end, it takes more time to un-packetize, decompress and turn it back into a phone signal. This process (minus the digitization and compression/decompression) is typical of all network transmissions. For typical data transmissions, latency isnt a major problem. But for phone calls, latency causes the annoying delay between when a word is spoken and when it is heard, much as it was in the early days of long distance calling. Callers end up talking over the top of each other in a manner that is disruptive and not tolerated in a typical business environment.

The following chart illustrates the latency problem (based on an article by David Passmore; "Delayed Voice over IP"; Business Communications Review; December 1997):

Process Time required
in milliseconds
G.729 compression 25
Buffer 40
PC-caused (I/O, IP stack, OS) 400
Inherent Internet latency 300
TOTAL 765

The threshold of delay that is not noticeable to the human is considered to be 100 milliseconds, and a traditional toll call meets this standard. Even assuming that all of the above latency factors are dramatically reduced, VoIP quality is not likely to reach toll quality any time soon, especially if the notoriously fickle Internet is part of the equation.

In a more recent article (Passmore, "No Ironclad Guarantees"; BCR; October 1998), Mr. Passmore argues that, "Guaranteed network performance in an IP network will require industrial-strength quality of service (QOS) features" in order to successfully handle an application such as VoIP. He discusses at length various QOS alternatives (Custom Queuing, IntServ Guaranteed Service, and DiffServe), opining that none of these solves the problem fully. He concludes that, " because there are no easy answers, managers may want to think long and hard about the basic premise: Should they really attempt to meet strict performance guarantees in their networks?"

Good question.

Reliability

Perhaps more significant than latency is the reliability problem. Everyones experience of the telephone is that, when you pick up the receiver, you get dial tone almost 100% of the time. In fact, the actual percentage of telephone service availability is 99.999%, the famous "five nines" of reliability. That averages out to 5 minutes of downtime per year; any more than that is considered unacceptable.

Data networks, on the other hand, are nowhere near as reliable. Data networks are routinely taken down for maintenance. Due to complexity caused by multiple protocols, multiple vendors, different service providers and platforms, various operating systems and network management systems, etc., it is virtually impossible for data networks to reach the reliability bar set by the telecommunications industry. The following chart (information taken from Mary Thyfault; "Voice-Data Integration"; InformationWeek; 13-Apr-98) indicates the average availability of various networks:

Availability Annual downtime
minutes days
Voice standard 99.999% 5.5 --
Private data networks 94% 31,536 22
Carrier data networks 91% 47,304 33
Public Internet 61% 204,984 142

The best private data networks are available about 94% of the time, on average, meaning a user can be without the digital equivalent of dial tone about 22 days per year. Compared to 5 minutes, 1 day per year sounds like a lot; 22 days down sounds well beyond reason. And we havent even mentioned the public Internet. It simply isnt good enough today, and it will be a long time before the Internet can handle business-grade VoIP traffic, if ever.

There are some who feel that toll call reliability is at least possible on the data side. Jason M. Beckett, Director of Sales and Marketing for Samsung Telecommunications of America states: "One-hundred per cent uptime can be achieved in the computer side of the industry, but" [you knew there was a but coming, didnt you?] " but not at a price most end users are willing to pay. The argument that the PBX is dead and all voice will be LAN-based will never be accepted by the customer." (Inter@ctive Week; October 1998)

While Mr. Beckett may be speaking from a telecommunications marketers perspective, we believe he has struck the right note: sure, a fully reliable data network can be built in theory, but it probably wont be because the cost is beyond what buyers are willing to pay.

Other Technical Difficulties

Though latency and reliability are the most significant challenges, there are many other technical difficulties confronting digital telephony adopters. These include:

When was the last time you had to re-boot your telephone?

Functional/Technical Requirements: The Bottom Line

There are indeed some possible technical and functional advantages offered by digital telephony, but the technical problems are currently unsolved and are likely to remain so. Traditional telecommunications work so well and so reliably that it begs the question: Why change now?

Change Factors: Improved Manageability

The management advantages of digital telephony should now be apparent. In particular, there is now a single network to manage instead of two. This means: one cable plant instead of two; standard network management tools can be used for voice and data; one network protocol is used on- and off-campus; perhaps even one institution-wide operating system; voice-related databases can be integrated with other databases; etc. From an organizational perspective, it can lead to a single, unified, streamlined IT department that handles all service and support for all things electronic.

Yikes!

Management Realities

Obviously, organizational changes, especially those involving "streamlining" operations and personnel, can be very difficult to implement and manage successfully, even when such changes are a really good idea. All kinds of political, budgetary, personnel and personality issues can get mixed in, leaving aside the technical training and support issues.

There are other management issues to consider as well. For example, IP telephony is not currently regulated by the FCC, hence the lack of access charges and other federal fees. However, IP telephony is under FCC scrutiny as was mentioned above. It is quite likely that the regulatory landscape will change dramatically over the next several years and could well change what today is a good idea into a bad (or at least non-profitable) idea tomorrow. In a related issue, some institutions have contracts with long distance providers that preclude providing any support for competing telecommunications alternatives; such schools are obligated to send all their long distance minutes over the providers network. IP telephony could be in legal conflict with such provisions.

Then there is the question of how to provide and pay for student services. Student long distance calling has long been a cash cow that produces "revenues" that can fund various telecommunications and other projects. If calls are sent over the data network, how does the institution recoup this revenue? Or is it simply lost? Remember, we are talking about several hundred dollars per residential student per year in revenue! If schools do attempt to recapture this income, how do they bill for it? The typical cost-per-minute model doesnt lend itself well to data communications.

Improved Manageability: The Bottom Line

Theoretically, managing one network is easier than managing two, and integrated systems and databases are easier to manage than dis-integrated systems and databases. Theoretically. However, convergence will also create a new set of management problems to deal with. In the end, convergence will ease some management issues, exacerbate others, and eliminate none.

Change Factors: Hype

As we stated earlier, the essential dilemma for the IT professional concerned with digital telephony is separating the hype from the reality. As the various quotes used throughout this article show, the hype is deep and getting deeper.

A telling example of VoIP hype was found in Business Communication Review ("Voice over IP Gateways: Sounding Good"; February 1998). This august publication proclaimed: "1998 is beginning to look like the year that packet voice could finally make it big." So far, so good. However, BCR goes on to ask the question: "Is the quality acceptable for business use?" "Yes", we are assured, "most of the time, under normal network conditions." And then, sounding a bit Clintonesque, the following is added: "Of course, this depends to a considerable degree on your definition of acceptable".

Without belaboring the point, lets just say that the phrases "most of the time" and "under normal network conditions" should be big red flags to network managers and telephone users everywhere. And when it comes to placing a phone call, "acceptable" means only one thing: you pick up the receiver, get dial tone, and complete the call successfully, every time.

Hype: The Bottom Line

Industry trade journals and vendors glossy brochures are rife with sentiments similar to the one above, although most leave the negative parts out. When it comes to digital telephony, it is clear that the marketeers are overhanging the market, which is, after all, what they are paid to do. Despite what is being said and written, it will be several more years before digital telephony "finally makes it big", if it ever does.

Conclusion

The following chart summarizes our perspective on whether it is time for IT professionals to make the move to digital telephony, based on the factors discussed in this paper:

Change factors
Time factor
now 5 years out
Economics no maybe
Functional/technical requirements maybe probably
Manageability no maybe
Hype always of course

Bearing in mind that, according to our equation, a successful technology decision depends on a combination of favorable economics plus improved functionality and/or manageability, things dont add up.

In the meantime, IT planners and managers can invoke familiar strategies to deal with the usual uncertainties they face while awaiting technology revolutions to arrive. These strategies include:

In summary, we believe that traditional PBX technology is proven, reliable, evolving, and here to stay. Digital telephony is still unproven and unlikely to overtake traditional telecommunications technology for the foreseeable future.

The end of the world as we know it?

Not quite yet.